As we probably all know: AI is currently pretty decent at making images. Of course most is easily recognizable slop, especially things produces from free-tier programs, and even most paid-tier services.

But it's not unreasonable to expect, or at least, imagine it gets better at it - in fact we probably should try to imagine a world in the not too distant future where AI programs really can make mechanically perfect artworks in any type of style. And not only images - for the sake of this discussion, we should posit it can do that with ALL types of media. Movies, books, comics, stories, music, games, architecture - even our favorite: Starcraft maps lol.

I'll say three clarifying things right away:

• 1. There IS some room for debate on whether that will actually happen or not - but that's not the point of this discussion. For the sake of this discussion, we should consider what it would be like in a world where AI programs really can _perfectly_ create pieces of art to arbitrary precision in whatever art style you ask it to.
  
  
• 2. For this discussion, by "mechanically perfect" I want to clarify I mean "IDENTITCAL to what ANY human could produce in that particular form of media." And so I mean: writings like Shakespeare, if that was the desire - but ALSO importantly: Not quite Shakespeare if _that_ is the desire; i.e. suppose you would object and say "perfection is essentially an oxymoron since it would lose those tiny 'human' imperfections which make art more valuable;" to which I want to respond with "consider AI can _intentionally_ make those tiny imperfections, so it has that 'human-like' quality." I hope that makes sense.
  
  
• 3. At the risk of sounding pretentious or something (especially since I'm also including SC maps in this definition lol) I'll just refer to all of that type of media as "art." So I'm not thinking only of images, or literature, etc - but basically, anything which traditionally has involved the creative processes of a person or people to any extent

So the central question is:

"What is the value of art in a world where AI can instantly create perfect art?"

I'll write my own thoughts:

Broadly I think there are a few things to consider when you want to think of value in art :

• A. In the final result - mechanically satisfying or aesthetically pleasing output. - basically, who/what made it or how it was made or when it was made is irrelevant.
  
  
• B. In the surrounding "story" or "history" behind the art.
  
  
• C. What does the piece of art say/imply.

Really I think C is in some ways central to the concept of 'value' and is not really at odds with A or B. Actually none are at odds with the others; these aren't all mutually exclusive things to consider.

To clarify my thoughts, I'll make one more point and then give an example.

Point:

You might have heard it said that a picture is worth 10,000 words. Let's take this as true. This essentially means that pictures/art are really saying something - even if it's just entertainment e.g. feel good video games, that entertainment inherently says something. (point C)

Take Tetris for example: You think it's not saying something? It is! It was made by a person, that person thought it would be worth their limited time on earth to program a fun little game, and so they did it - it inherently has meaning since it was made by a person and they sacrificed their time/put in the effort to create that thing (point B). It also has it's own tiny bits of the person's philosophies - "It's fun to push yourself to keep up with gradually increasing difficulty"

Also, Tetris is mechanically well produced (point A), especially for the time at which is came out. It has cute music, and simple/intuitive gameplay.

Now when it comes to AI:

I am essentially saying it can perfectly execute whatever value comes from point A. It cannot currently however replicate any value which comes from point B - at least on its own. Perhaps if you consider the person behind the machine doing the prompting you could tie that in with the output - that is an entirely valid position to take.

I order for AI to have value in its output which comes from concept B, it would have to have itself continuity (beyond whatever short context is in a chat window), and it would have to have I would argue true meaningful sacrifice put into its work. As it is now - there is _no_ sacrifice put in from the machine. Whereas when a human makes something - there is _always_ sacrifice to some degree.

In fact: in the past, when humans prodcue mechanically proficient works of art, a large part of why that is significant is actually from the fact that the person who did it poured themselves into it - they studied a long time, sacrificing their own time and effort in order to make whatever it is they made. It inherently had that sense of gravitas behind it. Even if the output was on the face of it 'stupid.' Look for example at that chinese movie, Niu Lai. At least for me, when I see those poorly CGI rendered images, I feel something which is coming solely from the fact that the creators poured themselves into it. If an AI made something like that, that feeling would be non-existent.

In the future by the way: I think it's reasonable to expect that in fact AI machines which are able to have perfect continuity in the same way that humans do, and accumulate experiences, and truly make meaningful sacrifices of themselves for certain things - then they also would be able to pour this sense of point B into whatever artwork they make. It might be slightly unrelateable to us as humans, but, not too far off. You could also imagine scenarios which have been considered already for a long time, which would draw this sense even closer to the feeling you get when humans make something. Take for example Terminator; you grow attached to him, and you even feel the weight of his sacrifice to protect John Conner at the end, because he had experiences and sacrifice.

As it is right now: I think that properly guided, AI can be a great tool for efficiency: and there is merit in just wanting fine output, e.g. playing whatever SC map you want right away, and having an AI make it for you. However, it can not replicate the inherent weight behind whatever a human makes - because a human is really saying something, making a statement, whenever they make _anything_.

Properly monitored, an AI can help someone make the thing they want to make, and make it say the thing they wanted it to say. It does lose the gravitas which comes from the person's willingness to sacrifice their time and pour in the effort to learn to produce the thing and do it well and to high precision - so it might feel less impressive; in fact, it _is_ I would say less impressive. But that's the way of things, I guess. They can regain that impressiveness, which also has merit, if they want to make it without the use of AI. That does in fact add some more value back into whatever is produced. So there's this sort of tension.

Finally: For AI mass produced, instant produced art: the substance of _what_ is being said is usually at least for now going to be 'slop' in the sense it will be the most basic or morals, the least interesting or introspective, and sometimes essentially nonsense. This is because again AI currently does not have life experiences; it doesn't right now produce novel insights which we can take away from - so basically, the point C of value; AI is sort of like just shooting a shotgun at it trying it's best in a scattershot type approach (this is my personal feeling). It might say something interesting, in fact - but whatever is said is not going to be rooted in a real person's experiences. In the future again this might change - AI artwork might truly convey novel, moving concepts and provide genuine insight into the 'human experience' or I guess 'experience of sapience.'

This means also that the use of AI in your own projects needs constant monitoring to make sure it's not introducing unintended meanings/statements/philosophies you don't really want to convey. From everything down to the minutes details, all that used to be truly touched by the people working on it; now, some of these things are being filled in with random 'words' from LLM, which even if mechanically proficient can take away from what you are trying to make.

^

Aight, that's a lot of my thoughts, but I thought I'd try to open up the topic for discussion - Idk if anyone else is interested in this stuff, but I think it's becoming more and more relevant. I would love to hear people's thoughts on it all

Post has been edited 2 time(s), last time on Sep 7 2026, 5:29 pm by Zoan.

I think I would respond to this as if Sartre might.
Sartre would argue that we continually define and create ourselves through our choices.
If AI writes your poem, paints your picture, or shapes your opinions, it raises the question of whether the final product truly expresses who you are or simply reflects your choice to rely on AI.
I think AI makes us less authentic and failing to take responsibility for how we use AI definitely makes us less authentic and has no value.
In this sense, the value of art is not simply in how perfect the final product is, but in the authentic human choices and intentions behind it.
If those choices are completely handed over to AI, the work will lose its personal meaning and value because it no longer represents the individual's authentic expression.

My biggest fear with AI is its capabilities as a brainwashing tool. Any media produced by AI will inherently carry bias attuned to the AI creator's will. If mechahitler controls the art, than art criticizing mechahitler will cease to exist.

The human element is important to me as well. I'd rather support someone making something out of love than something churned out by a corporation - this isn't exclusive to AI, but it's the same principle. I'd rather feed a hungry artist than give the ruling elite an extra dollar.


That sacrifice is forced on ordinary people, though.

I'm not entirely anti-AI, but I don't use it myself. I'll paint with my own blood, not the blood of the world!

It's not possible for an AI to create anything "new". You even use Shakespear, a human, as a reference in your arguments. A human can write works like Shakespear or not like Shakespear without having ever heard of or learned the works of Shakespear. The LLM requires all of Shakespear's works and all spinoff works as input and only produces anything within the parameters it was trained on.

"AI" is just a black box that gives an output for every input. There is no thought process, there is no lived experience, no emotion, and it can't produce anything outside of the parameters it was trained on. Because anything an LLM trains on becomes numeric soup with P-value hacking, anything it regurgitates sits right in the uncanny valley and nothing is cohesive. It merely follows the patterns it was trained on.

"Perfect" is being thrown around a lot, and you proposed an argument about human imperfection or whatever, but you're forgetting that AI is not - and never will be - "perfect" either. The mechanism which allows AI to generate believable content is the same mechanism which allows it to hallucinate misinformation or even glitch out entirely when the inputs deviate too much from what it's been trained on. An intentional accuracy target of lower than 100% is how these can function at all, otherwise AI could only produce anything if it was given an exact input that it was trained on. So if your goal target is to "perfectly match a human" you still can't do that with 100% accuracy, it is impossible by design.

There's some mention of "guided" and "monitored" but AI is just an algorithm that produces outputs from inputs, there is no "guiding" it because it's not a "being", it is a machine compilation created by humans and human input.

I have personally never seen or heard of any AI art that was "good", either.

Ultimately, I think machines are meant to do manual labour so people could focus more on art and we'd all have UBI, but right now it's backwards. Too many disadvantages of pollution, water consumption, hardware scalping, the deferral of creativity and thought actually making people dumber, IP theft, and the greatest corporate double standard that has ever existed. People got fined for downloading music or e-books, but AI companies can do this AND profit off of it either without penalty or a slap on the wrist. Big corpos are axing artists for AI and charging more for their products to make line go up in the biggest race to the bottom of our time.

I second Heinermann's post. It's pretty much 1:1 with my views on generative AI.

The environmental impact and ethical concerns outlined in the final paragraph are chiefly why I consider myself to be against gen-AI. I otherwise wouldn't really care at a surface level if someone wants to use gen-AI to "create" "their" product. I would however still look down upon it in the same way that I look down upon e.g. traced artwork, stolen melodies, and plagiarized writing; as a product lacking real or unique value. I would argue that plenty of gen-AI users themselves look down upon it, as evidenced by their attempts to claim the output as being produced by their own hand rather than AI-generated, and in the case of visual art going as far as faking the creative process (sketches, WIP snapshots). If they are not willing to be honest about their work, how can they expect it to be treated as such?

Personally I would probably be open to using gen-AI to lightly inform my output or as a source of inspiration, but its use in its current form goes completely against my values and I will not compromise, especially when abstaining takes absolutely nothing since it's nothing more than a luxury.

R.I.P Aaron Swartz

Post has been edited 1 time(s), last time on Sep 8 2026, 2:54 am by DarkenedFantasies.

"Generative AI is the asbestos of the internet."

Saw that quote a while back and it's pretty on point imo. I'll probably repeat some of the points brought up by Heinermann and DarkenedFantasies, but I just want to get it out.

I've grown to hate gen-AI with a passion. I'm all for using AI for stuff like medical science, astronomy and things like that, which saves scientists loads of time sifting through data looking for things, but using it for generating media comes with so many problems.

• It starts already before anything has been created; the models are largely trained on data without the original creators' consent. It's basically piracy on an industrial level. Hopefully the (supposedly) ongoing lawsuits will kick the bastards back into the gutter where they belong.
  
• Generated photorealistic images or video is already a huge problem when it comes to misinformation, fraud and much, much worse things.
  
• (Dis)Services like ChatGPT seem to make people stupider. Many people who use it a lot appear to no longer be able to find or process information on their own, cause they're either too lazy or they've forgotten how to do so. This is the reason why I like to call gen-AI fascism's best friend.
  
• AI-generated "art" is soulless. Art, whether it's drawing, literature, music or film, is meant to show the artists' passion, emotions and hard work, not to be churned out and thrown away after it's been partially consumed.
  
• The environmental impact also needs a mention, given the huge amount of data centres popping up everywhere. While you can't blame AI for all of them, it's certainly a large portion of them.
  
• Hardware prices have also dramatically increased because of AI. I bought a new computer in October last year, right before the prices went insane. Checked the RAM prices two months later and they had tripled. Same with a HDD I ordered with it. It's the crypto-mining phase all over again, but worse.
  

Honestly I'm just sick of the whole thing and I sincerely hope the whole industry will crash and burn soon.

I have some differing viewpoints.


I disagree, if you accept some premise I'll highlight below.

At the same time, I'll also try to address another point mentioned:


First off: I am aware this is a hot topic, deeply related to the notion that many prominent LLM's training data has been taken unauthorized.

I am identifying two statements both of which I want to address:

Point 1. AI took intellectual property unauthorized
Point 2. AI's output is just a derivative / amalgamation of all its training data.

I will start with point 2.

Point 2:

I think this statement is correct; we all know how LLMs work, I trust, so no need to dwell no that.

But what I would add is that: what evidence do you have that humans aren't doing the exact same thing, just on a more specificed level than LLMs currently do? I believe actually that (aside from souls etc, so on a purely physical level:)

Claim: Humans really are just outputting via a series of complex interactions of neurons in our brains an output which is a result of the input into us: our own "Context Window" which is the sum total of our ENTIRE life experiences. This is essentially the same thing that LLMs do.

Consider that position, then. Then, essentially, point B is really only apparent so far in LLM progress as a negative because right now the output is just 1. not different enough, since we have not worked out all the kinks so far and 2. not based on enough life experience/training data.

So, if this viewpoint is correct, given enough 'compute' and training data, AI 'processing of inputs' (synonymous in this viewpoint with just the word 'thinking') would be indistinguishable entirely from a human's output given the same data and inputs.

Further, if this viewpoint is correct, then we could not/should not label their output as plagiarism unless we want to label ALL creative works not done in a vacuum as also plagiarism, which we obviously don't

Roll with me for a second and suppose that's true; from that viewpoint, I'll say something about point 1.

Point 1:

First off, there are instances from the leading companies that they took/gathered data unauthorized/illegally. Obviously, this is wrong. In the simplest abstract viewpoint "Company stole stuff, used it to make stuff" <- that is clearly indisputably wrong.

Modulo that we can delve more into the other side of the issue raised of LLM that all their output is essentially some form of plagiarism, point 1.

Let's take on the previous viewpoint for a second, that their thought processes are some form of exactly what humans do. In this sense, we should ask the question:

Is it wrong for e.g. an author to read another author's book?

If they do, surely on some level they will be influenced by whatever it is they read; and when they write their own works it will have influences from their 'training data.'

Of course, there is some obvious threshold of what constitutes plagiarism and what constitutes just inspiration. But why can't we just take the view that LLMs are inspired by all the works which went into training them, rather than that they are just plagiarizing them? It's not as if they literally output the same verbatim works which went into them, right?

__

As for the environmental concerns: those may be valid. Personally, I am at first not really convinced on any of its reality. I have only looked into the water argument, and to me it does not seem that compute centers require inordinate amounts of water. All things I've seen are just fearmongering and when you look into it the numbers are just claimed projections, no hard actual real data, nor put into realistic perspectives compared to water usage of literally every other industry.

However: That is in my opinion a separate topic of conversation. I would want to separate that train of thought from the philosophical questions about AI and Art and things like that that I mentioned.

On the subject of training, plagiarism, and inspiration, what makes the difference to me is that between humans it's an exchange of information that is mutated by individual experiences. Whereas generative AI is an aggregate of these experiences. It can be argued that nothing is completely original, but I see it as a gradient where either end cannot ever be fully reached:
[ Original---Inspired---Imitated---Plagiarized ]

So it could also be argued that on that same token nothing is completely plagiarized, as the plagiarist still had a minimum of input into it, in the effort put into plagiarizing the work and what-not. It's just not the kind of effort that we consider to be valuable.

When someone plagiarizes someone else's work, the information hasn't been mutated into something new, only the attribution has been changed, and the perpetrator either aims to gain something they haven't done any of the work on (thief), or is engaged in the act of imitation (copycat). In either case, we view this as a fraudulent act (to varying degrees) because broadly speaking it offends our sense of justice. We're fine with it when someone (legitimately) comes up with the same thing independently because it's an accident, not intent.

The issue of plagiarism with generative AI stems from a core aspect of the work (the training) being taken from the public and used for private gain; the attribution of the work has been claimed or hidden, and the broader public had their work used without permission or compensation. So to respond to the following quote:
It appears it does, as poorly or insufficiently trained models have been shown to output very similar or nearly identical work to what was used to train it (as claimed by artists who have spoken up after recognizing the matching aspects between their work and the AI-generated product).

It simply becomes less evident as a wider range of input blurs the pattern into something we can no longer easily match up to the original work, which seem to make the generative output inch closer towards the center of my scale above, away from pure plagiarism. Whether that is truly the case, or just an illusion caused by our pattern recognition being deceived, I'm not sure. I don't think anyone would make the argument of plagiarism against someone who has trained their model strictly on their own work, though.

I believe the accusation of plagiarism is a simplification that encompasses the unjust application of other people's work, and that these models or the users don't practice subjective modifications of the work that's been trained on like an artist would (aka "soulless"). The creativity comes from the artists they continue to loot from; the model only autofills the prompt without any additional modification, and the semblance of originality is but a product of random generation.

Oh well that is a problem. I don't think it does that too often if at all anymore? In the earlier models I would guess it did that?

But in the view I put forward earlier: that would be akin to a mentally deficient person trying their hand at writing, and their attempt ends up being just copying essentially verbatim someone else. (in this view early LLM = mentally deficient person) People in highschool end up doing that all the time, that's why they learn about explicitly what counts as plagiarism and what doesn't.

It is not acceptable on an adult level, especially for something you market or sell; but I think for the most part LLMs are past that, and the trajectory of them seems to be such that it will properly only be inspired by its training data rather than directly copy it in the future.

Again there is some varying assumptions in this whole discussion:

1. Where LLMs are now
2. Where LLMs are going in their future trajectory
3.What an idealized LLM would be like to work with and the implications of that in art and stuff etc.

Of course not.

An LLM does not "read". Even if the human brain is a glorified flesh LLM, the methods we use to interact with reality are inherently different. To me inspiration comes from emotion.

Maybe we could hook one up to some kind of dopamine receptor. Really push the ethics on this stuff. Give one a pain receptor to see if it makes it a better artist!

Materialism vs Idealism/Dualism
My suspicion around a lot of the angst around AI generated art actually comes from a divide that has existed long before LLMs. Materialism vs Idealism, or Materialism vs Dualism. The idea that there is something 'special' or even 'magical' about the human mind, and thus all this, what I would call, wishy-washy thinking around art, all these nebulous barely defined concepts around it. Elevating it to more than it is.

Anyway, that's such a massive ideological debate to dive into I'm not even going to attempt it - just want to raise that's my big suspicion what is underpinning a lot of what is going on in these AI vs anti-AI culture wars.

AI in STEM vs AI in 'The Arts'
Another thing I see a lot, and I want to call out as being pure arbitrary BS, is people who hold these quite frankly hypocritical opinions about how they're perfectly fine and supportive of AI being used to advance fields in science, medicine, mathematic but then have complete rage-outs, moral judgements flying everywhere, when it comes time to using AI in arts.

News flash - creativity doesn't adhere to your meaningless arbitrary boundaries. The AI discovering a new proof in mathematics or helping scientist craft some new drug is exact same process as the AI helping an artist with their workflow in creating their painting or composing their music or whatever.

Either decry both, or decry none. You can't have it both ways without being a total shameless hypocrite - IMO!

The Process vs the End Result
This a cause of the divide above, STEM is almost always focused on the end result. No one gives a fck about how much blood sweat and tears you poured into your hypothesis. They care if it's valid or invalid. Your cancer cure either works or it doesn't. But art there are far more many people with this obsession with "the process", van Gogh cutting off his ear etc you know what I mean. It's not just the end result. LLM and generative AI is rapidly diminishing the process and fast-forwarding straight to the end result, and the process crowd don't like that.

P.S. We saw this exact panic happen when photography first came into being. Go read your history sooo many people were crying out 'this is the end of art'. No. News flash. That didn't happen. We just now have a new form of art that existing alongside the existing form. Did a lot of painters lose their job? Yes. But the world kept turning. So it shall be with AI. Accept it.

My reply to the process-obssessed people (I am not one of them) is calm the F down. You guys aren't going anywhere. Go spend 50 hours crafting your painting while I type into my prompt and get it in 5 seconds. You do your thing, and I do my thing. If they just accepted that, they would be happy. That's why I suspect some of them are actually lying. What they are really crying about is the fact that there is now alternatives that will cut into their bottom line. The 'end result' crowd will discard these people immediately (as I have) because we do not care about the process. We want the end result as cheaply and efficiently as possible (preferably without any quality drawbacks of course).

Breaking down the nebulous term "Creativity"
We're in Serious Discussion, so let's get serious. I'm seeing a lot of wishy washy generalisations around this word in the context of the AI debate, and it's getting us nowhere. If you actually research creativity itself, a topic that has existed long before the LLM debate of course, you will see there are multiple types, and AI can do some but not others. Let's break it down. Rapid fire, and you can do your in research to go more in depth if you want.

Margaret Boden's model:

Combinatorial Creativity - taking two familiar but unrelated concepts and mashing them together to create something entirely new. This is the most common form of human creativity. It is also the type the AI is most capable of. So when the smooth brains decry AI as 'only' being able to do this type, they're also decrying the vast majority of all human art... I wonder if they realise that.

Exploratory Creativity - This is where someoen takes an existing style or set of rules and pushes it to its absolute limits to see what hasn't been done yet within that space. Think of it like min-maxing. Example is constantly modifying a compound systemically one bit at a time to discover what changes in the drug's properties, maximising effectiveness while minimising side effects. Again, AI is a massive accelerator of this type of creativity. It can be given the parameters of the task and then iterate over and over again, fully mapping out 'exploring' every molecule, every conceivable variation of a melody, or image style.

Transformational Creativity - This is the rarest type of creativity, when someone completely breaks or alters the fundmental status quo, creating a paradigm shift. Example when Picasso invented Cubism, shattering the rule that paintings had to represent a single realistic perspective. The LLMs/generative AI cannot do this and will never be able to do this in its current form. Most humans never achieve this in terms of culturally (inventing cubism for the first time - once it's invented, it's invented) but every human typically does this just in the normal course of growing up (eg a child independently discovering they can make their blue and yellow crayons to make green).

Now that is just one model of creativity. There are other models I encourage you to explore like the Four C model by Kaufman and Beghetto, the Four Ps model by Mel Rhodes, and others I may be ignorant of, and in particular what are AI's strengths and weaknesses in relation to them. I simply wanted to make the point that people are being FAR too simple-minded in their discussion of creativity as if it is some single monolith. No. It's not.

My own positive experiences with AI
Now there are a lot of theoretical and abstract arguments for or against the use of AI, and I want to close by taking things down to earth and showcasing my actual lived experience of how much AI has improved my own hobbies and life. I am fully pro AI, I think it is making humanity better, not worse, and I am feeling the benefits myself every day.

ART GENERATION
Now I work on the mods Legendary Lore and Victory Conditions Overhaul on Total War Warhammer. These are some of the most subscribed mods on the workshop with hundreds of thousands of subscribers. AI has levelled me up immensly here in multiple ways.

For starters art. My goal with the Legendary Lore mod is for every Legendary Lord and Legendary Hero to have art go along with its text popup. Now many never had any art done, not even fan art, and so I used to resort to just a generic art instead. I also commissioned art, which was costly, $300 USD usually, I did about a dozen of these.

Other pieces were only in black and white (I wanted full colour) or were potrait (I needed landscape and had to do bad crops), or were mismatching art styles or art styles I just plain thought were ugly/bad.

Then along came generative AI (it was initially shit and I didn't use it, but then end of last year it suddenly got REALLY good) and now suddenly I do not have this problem. I can make bespoke art for every single Lord. I can even fix bad crops if I wanted to preserve the initial piece. It cost me basically 0 dollars and 2 minute to generate. If I didn't like the generation, I could generate again. I actually threw several of my commissions in the bin, because the AI literally generated superior results. This was a total win for me, and I'm never going back.

Sometimes people say I 'stole' the art. I just think they are using a very disigneous usage of that word. How can I "steal" something that never existed until I created it via the combitorinal creativity machine of the AI? It's a stupid argument. If you want to make the argument that the LLM companies stole the art and fed it and used it to train their machine illegally, fine, I won't fight you on that, and the courts are settling that right now. Maybe it's just a matter of ensuring proper compensation. But to say the thing I generated that never existed before is somehow 'stolen' is just completely wrong. Trust me, if it existed, I would have already been using it...

CREATIVE WRITING
Legendary Lore is full of fan fic. I used to manage contributions of other authors, proofing their work. When AI first came out, I hated it, because some people started flooding me with generated crap that was terrible to read and basically unfixable. But then, same time last year, suddenly the AI writing started to get good. It has gotten to the point now where I can no longer tell, or there rae only extremely subtle tells you know as someone who has their face in it every day. Again, this is a big win because creative writing was a laborious, very time intensive process, something often players didn't care about (but I did because of my own personal standards) and now I can write better AND quicker for my flavour texts. It's a win-win.

CODING
I have no formal training in coding. All I ever did was Starcraft triggers. But in Total War modding I needed LUA to achieve a lot of what I wanted to do as a modder. So I slowly learned and got better, but then along comes AI, and not only is it a great teacher, accelerating my learning and explaining code I show to it trying to figure out exactly what's happening logically, to it also being able to code. If you aer a programmer you know there is a lot of coding that is like 'shut your brain off coding' where you just need to do some laborious time consuming thing. Now I could hand that off to AI. Again, it was a massive producitivty BOOSTER. No quality trade-off. IF there was a bug, 99% of the time it was my own fault, my own flawed understanding, or the AI's lack of context and me not catching it. Not only that, but AI could also code stuff I hadn't even attempted to code yet. So it allowed me to 'punch above my weight' as an amateur programmer BIG time.

GENERAL KNOWLEDGE
I actually have a lot of fun just hitting the AI with any shower thoughts I have. Any concepts I've always wondered about, but could never find answer to, etc. Yes, I know hallucinations can happen, but usually I have a positive experience and walk away mentally richer, not poorer.

So there you have it. Thanks for reading my essay. In summary, AI is awesome, incredible tool, future is bright. I strongly believe all the haters and doubters are on the wrong side of history, and just have flawed moral reasoning around this issue, possibly due to a non-materialist mindset, a process rather than end result mindsight, or simply selfish 'I don't want to lose my job' mindset.

Sorry for double post I thought it best if I put my general thoughts in one post then my replies to people in separate post.

Yes Zoan I agree with you about your general concepts here, how LLMs and humans haev a lot in common, in fact they even tried to modelled the connections directly like neurons. There's a couple key differences. One is static vs continious learning - we constantly updating our worldview in real time. The AI's undersatnding is frozen as soon as its training run completes.

The ethical issue, I believe even if you accept the LLM similarity thing, the ethical issue comes from the industrialisation, NOT from any core difference. No one gets angry when a human digests perhaps a couple thousand books in their lifetime and then outputs a novel or two.

But when LLM does the exact same thing, the scale is completely different. The mega corpo scrapes every single human book, feeds it into the machine, now it's pumping out 1000 novels per minute.

The outrage comes because this is basically cheating or hacking the pre-existing system, all that time and effort from all those people, now taken for free by some mega corp and directly competing with them now, and the money funnelling into this small group of people. It's fundamentally unfair.

Even if you took money out of the equation, let's say they were purely non-profit and gave all those books away, I think there is still an element of unfairness because of the attention economy. The humans with their measly novel output and measly lived experience cannot possibly hope to compete with the ultra-mega experience of the LLM and its massive unbeatable output.

This is a misundersatnding of hw machine learning actually works. LLMs do not calculate statistical significance or hack P values. They use gradient descent and backpropogation to map concepts into high dimensional mathematical space.

A p value is a tool used in statistics to say 'is there enough evidence to prove this drug works or was it just a lucky coincidence?' its about testing a hypothesis.

LLMs do not test hypotheses. They do not care what is 'true' or 'statistically significant'. They are purely optimisation engines. By dismissing it as 'numeric soup' you're completely ignoring the incredible breakthrough they've achieved of proving they can map complexities of human language and reasoning onto geometry.